Towards a theory of how the structure of language is acquired by deep neural networks

Open in new window